Oecologia DOI 10.1007/s00442-006-0447-7

COMMUNITY ECOLOGY

Space invaders? A search for patterns underlying the coexistence of alien black rats and Galápagos rice rats

Donna B. Harris · Stephen D. Gregory · David W. Macdonald

Received: 19 September 2005 / Accepted: 26 April 2006 © Springer-Verlag 2006

Abstract The introduction and spread of the black rat increased with R. rattus density perhaps reXecting an Rattus rattus is believed to have caused the worst increase in foraging eVort necessary to compensate for decline of any vertebrate taxon in Galápagos. How- the costs of interspeciWc exploitation or interference ever, the “extinct” Santiago rice rat competition. The distribution, microhabitat selection, swarthi has recently been rediscovered in sympatry and abundance–habitat relations of N. swarthi suggest with R. rattus providing the Wrst exception to this gen- that the endemic cactus O. galapageia may facilitate eral pattern of displacement. We carried out an explor- interspeciWc coexistence. Further research should atory investigation of this novel system with the aim of include a comparison of inter-seasonal resource prefer- identifying patterns that may facilitate the apparent ence and foraging activity of the two species coupled coexistence of the two species. We carried out an with replicated Weld experiments to conWrm and quan- extensive survey of Santiago Island to map the current tify competition and to elucidate the mechanism of distribution of the endemic rice rat and to explore competitive coexistence. broad scale distribution–habitat associations. We then used live-trapping, radio-tracking, and spool-and-line Keywords Cactus · Introduced species · Microhabitat · tracking to quantify abundance–habitat correlations Spatial segregation and to test for evidence of interspeciWc spatial segrega- tion, alteration of N. swarthi activity patterns (spatial and temporal), and microhabitat partitioning. We Introduction found that N. swarthi has disappeared from part of its historical range and appears to be restricted to a 14 km The patterns underlying the coexistence of ecologically stretch of the north-central coast, characterised by high similar species have been a major focus of community density of the cactus Opuntia galapageia. In contrast, ecologists for the last 40 years. Much of this research the generalist R. rattus was found at all survey sites. We has concentrated on small communities and found no evidence of spatial segregation, and home there is now ample evidence to show that spatial, tem- range size, temporal activity and density of N. swarthi poral and resource partitioning play crucial roles in the did not vary with local density of R. rattus. However, structuring of these natural communities (e.g. Scho- pre-dawn and post-dusk N. swarthi activity levels ener 1974; Price 1978; Abramsky et al. 1979; Bowers 1982; Morris 1987; Kotler and Brown 1988; Jorgensen and Demarais 1999; Jones et al. 2001). In contrast, the Communicated by Hannu Ylonen role of interspeciWc competition and coexistence in the restructuring of invaded communities has received rel- D. B. Harris (&) · S. D. Gregory · D. W. Macdonald atively little attention. Although many authors postu- Wildlife Conservation Research Unit, late competition with exotic small to explain Department of Zoology, University of Oxford, Tubney House, Abingdon Road, Tubney, Abingdon, OX13 5QL, UK the demise of native species (e.g. Brosset 1963; Alvarez e-mail: [email protected] and Gonzalez 1991; Stephenson 1993) much of this 123 Oecologia evidence is circumstantial and commendable studies of ever, the rediscovery of N. swarthi in sympatry with R. competitive processes in invaded systems remain rare rattus on Santiago Island is an exception to this rather (e.g. Gurnell et al. 2004). This is unfortunate consider- clear pattern of alien arrival and native . This ing the potential contribution of such research to the is particularly intriguing when we consider the chronol- Welds of community ecology and evolutionary biology ogy and geography of the R. rattus invasion. Genetic (Yom-Tov et al. 1999; Shea and Chesson 2002; Cour- analyses suggest that R. rattus Wrst landed ashore at champ et al. 2003) and the provision of information rel- James Bay, Santiago in the late 1600s (Patton et al. evant to the conservation of endangered species (e.g. 1975). Subsequent, separate introductions to the archi- Ligtvoet and Van Wijngaarden 1994; Krupa and Has- pelago, resulting in the current colonisation of 33 kins 1996; Macdonald et al. 2001; Zavaleta et al. 2001; islands (Charles Darwin Research Station (CDRS), Bryce et al. 2002; Courchamp et al. 2003). unpublished data) and successful invasion of all Galá- An unexpected opportunity to explore such a sys- pagos habitats, coincided with the loss of Nesoryzomys tem has arisen between endemic and intro- and species wherever the black rat became duced black rats in the Galápagos Islands following the established (Clark 1984; Key and Muñoz Heredia 1994; rediscovery, in 1997, of a population of the endangered Dowler et al. 2000). So, what conditions may foster the rice rat Nesoryzomys swarthi on the north-central coast coexistence of N. swarthi and R. rattus on Santiago? of Santiago Island (Dowler et al. 2000). This species At the coarsest level, species coexistence may be had been presumed extinct since its initial collection at facilitated by diVerential macro-habitat selection (e.g. Sullivan Bay (Fig. 1) on the north-east coast of Santi- Abramsky et al. 1990; Bryce et al. 2002). At a Wner ago in 1906 (Orr 1938). It was believed that N. swarthi scale, where species are sympatric within the same hab- had met the same fate as at least half of the species of itat, spatial segregation may be maintained through the once diverse assemblage of 12 endemic spe- interspeciWc territoriality (e.g. Lofgren 1995). Alterna- cies; the Galápagos rice rats (Orr 1938). This group had tively, microhabitat segregation may act to partition dominated the mammalian fauna of Galápagos but space (e.g. Bowers et al. 1987) and/or resources (e.g. since the discovery of the archipelago in 1535, this Price and Waser 1985; Monamy and Fox 1999). A shift group has experienced the highest extinction rate of in temporal activity patterns may occur to minimise the any vertebrate taxon in Galápagos (Clark 1984; Dow- probability of contact with the dominant species (e.g. ler et al. 2000). Today, just four endemic species Ziv et al. 1993; Jones et al. 2001). remain: N. narboroughi and N. fernandinae on Fernan- In our introductory survey of this novel system we dina, N. swarthi on Santiago and Oryzomys bauri on used a combination of radio-tracking, spool-and-line Santa Fe (Dowler et al. 2000). Circumstantial spatio- tracking and live-trapping to test two preliminary temporal evidence suggests that the introduced black hypotheses: (1) Given the strong case for displacement rat (Rattus rattus) played a leading role in the loss of of other Galápagos rice rat species following R. rattus the Galápagos Nesoryzomys and Oryzomys species establishment across the archipelago, we hypothesised (Brosset 1963; Niethammer 1964; Clark 1984). How- that R. rattus and N. swarthi would exhibit segregation

Fig. 1 Map of Santiago show- ing hair tube survey sites. White symbols indicate sites occupied by N. swarthi and black symbols indicate unin- habited sites. All sites were occupied by R. rattus

123 Oecologia in macro- or micro-habitat or in space per se. (2) We distributed among a shrub layer that mostly consists of hypothesised that N. swarthi would respond to increas- Clerodendrum molle, Castela galapageia and Lantana ing R. rattus density by adapting home range use or peduncularis. The area is exceptionally dry as it lies in temporal activity pattern to minimise interspeciWc con- the rain-shadow of the larger Santa Cruz Island and the tact and/or maximise resource acquisition. highlands of Santiago. Past search eVorts for N. swarthi To validate the implications of our Wndings for the have focused on southern and highland regions (Clark conservation status of this endangered, endemic spe- 1984; Dowler et al. 2000; CDRS, unpublished data) but cies it was of paramount importance to begin by delin- have failed to Wnd any N. swarthi in these areas. Previ- eating the distribution of N. swarthi in relation to that ous research has demonstrated that R. rattus is noctur- of the exotic R. rattus. This was achieved through an nal, exhibiting some activity in each hour of darkness, extensive census of the previously unexplored parts of reaching an activity peak 2–3 h after the onset of dark- Santiago. The investigation then proceeded to focus ness and that this pattern appears to be fairly general down through progressively Wner spatial scales from within the species (Barnett et al. 1975; Meehan 1984; habitat associations to a detailed contrast of microhab- Tobin et al. 1996; Innes 2005). Both Galápagos black itat selection and a brief exploration of activity pat- rats and rice rats are omnivorous including a variety of terns. fruits, seeds, invertebrate matter and carrion in their diet (Clark 1981; Jackson 1993). R. rattus is larger than N. swarthi (mean mass of male adult 183 and 115 g, Materials and methods respectively and female adult 141 and 91 g, respec- tively) (D.B. Harris, unpublished data). The data for Study site and species this study were collected over 3 years from 2002 to 2004. The volcanic Galápagos Islands are situated 960 km west of mainland Ecuador in the PaciWc Ocean. The Current distribution uninhabited island of Santiago is the fourth largest Galápagos island at 585 km2 with a maximum elevation To delineate the current distribution of N. swarthi, of 907 m (Jackson 1993). The climate is strongly inXu- baited hair tubes were used to survey for the presence enced by oceanic currents and there are two main or absence of N. swarthi and R. rattus at 22 sites pre- seasons; the warm/wet season (January–May); charac- dominantly located on the previously un-surveyed terised by warm temperatures with variable and often north coast (15 Arid Zone sites) of Santiago. We also heavy rainfall at all altitudes, and the cool/dry season sampled inland sites in the Arid Zone (3 and 4 km (June–December); characterised by cool temperatures inland), the Transition Zone (5 and 6 km inland) and with constant light rain and mist in many parts of the three highland sites (one Transition Zone, one Zanth- highlands but almost no precipitation in the lowlands oxylum Zone and one Fern-Sedge Zone). Each hair (Mcmullen 1999). The mean annual rainfall in the tube was constructed from a 30 cm length of 7.5 cm Santa Cruz arid zone for the period 1965–2004 diameter PVC tubing that was bisected longitudinally. was 491.44 mm, notably higher than the median of The bottom half was nested into the top and bound 277.55 mm, ranging from 63.6 mm in 1985 to together with wire. The top and sides of the tube were 2,768.7 mm during the El Niño of 1983. The tempera- lined with double sided carpet tape. The bait (peanut ture is comparatively stable with an annual mean of butter and oats) was twisted in a piece of muslin the 23.95°C§0.14 SE ranging from 22.6 to 26.2°C (CDRS, ends of which were pinched between the two tube unpublished data). The vegetation of the Galápagos halves. Tubes were wedged between or tied to rocks islands has traditionally been classiWed by altitudinal and vegetation at ground level. Thirteen tubes were zone. In order of increasing altitude these are the Lit- placed at each site in a T-shaped transect (7£7) at 30 m toral (Coastal) Zone, Arid Zone, Transition Zone, intervals (consistent with the trap spacing—see below). Scalesia Zone, Zanthoxylum Zone, Miconia Zone and At the coastal sites, tubes were placed 100 m back from Fern-Sedge (Pampa) Zone (McMullen 1999). The San- the shore to prevent bait consumption by hermit crabs. tiago rice rat (N. swarthi) is currently known to exist as Tubes were collected after 1 week. The tubes were one localised population situated on the north-central soaked in warm detergent solution overnight to loosen coast of Santiago, in the Arid Zone. This region is com- the hairs. Hairs were extracted with Wne forceps and posed of typical Arid Zone vegetation; the cactus dried on tissue paper. PVC glue was smeared onto a Opuntia galapageia and primary tree species Bursera microscope slide and hairs were laid carefully on to the graveolens, Cordea lutea and Croton scouleri are wet glue. After the glue had dried, hairs were carefully 123 Oecologia extracted with Wne forceps, discarded and the hair trapping was carried out from December 2002 to imprints were identiWed to species by individual cuticle January 2003. patterns according to reference slides (Teerink 1991). As limited resources precluded repeat surveys at diVer- Spatio-temporal activity ent times of year, tube visitation likelihood by N. swar- thi was maximised by surveying in August (2004), a Radio-tracking was used to investigate diVerences in predictably dry time of year (Snell and Rea 1999) when N. swarthi home range parameters and activity pat- N. swarthi are likely to be food limited and R. rattus terns in relatively high and low R. rattus density areas. numbers are in decline (Clark 1980). Preliminary live-trapping at multiple sites was Wrst undertaken to identify two areas of similar N. swarthi Habitat associations density but one with low R. rattus density and the other area with a relatively high R. rattus density. The abun- Habitat surveys were carried out at all of the coastal dance data were based on estimates from 10 days of hair tube sites during the survey to determine the habi- trapping on grids of 49 collapsible Tomahawk traps. tat type associated with the presence of N. swarthi. Habitat surveys were conducted in each area and com- Abundance–habitat correlations were quantiWed by pared prior to the radio-tracking phase to test for live-trapping on eight trap grids in the area of known underlying variation that may aVect home range use. occupancy, in the vicinity of a location locally known as The densities of C. molle (the dominant shrub), shrubs La Bomba (0°11Ј10ЉS, 90°41Ј59ЉW). Four of these grids (all species including C. molle if this was the nearest were located on the coast (100 m from the shoreline) plant), trees (all species) and cacti (O. galapageia) were and the other four were situated between 1 and 2 km compared between grids by measuring to the nearest inland. The T-square method (Greenwood 1996) was individual of each of the four plant types from 20 ran- used to obtain an index of mature cactus density, tree dom points on each grid and comparing distances using density and shrub density for each hair tube site and t tests. Mean C. molle and mixed shrub size were com- for each grid. The species of trees and shrubs included pared by measuring the volume (maximum in the density estimate were recorded. Cactus, tree length £ width £ height) of each of the 20 plants in the and/or shrub density could not be recorded at sites sample and comparing volumes between grids using t where the point–plant and/or inter-plant distances tests. Mean tree size was compared by measuring the exceeded 70 m as the rugged terrain precluded reliable diameter at breast height (DBH) of the two dominant judgement of the next nearest plant in many cases. tree species—B. graveolens (t test) and C. scouleri Lava structure was coded within a 2 m radius plot at (Mann Whitney U test). The species ratio was also each of the survey points within which lava was classi- compared using chi square analysis. Cactus size struc- Wed as either predominantly Xat or broken/cracked. ture was similarly compared by classifying the 20 cacti Habitat surveys were conducted at Wve of the hair tube according to the number of cladodes [0–25 (small), >25 points (90 m spacing) and at 20 randomly selected grid (large)]. Other potential sources of environmental het- points on each of eight trap grids. In addition, the rela- erogeneity, for example, rainfall and sea spray, were tive proportions of dominant plant species (i.e. plant minimised by the close proximity of the two grids species composition) within the tree and shrub samples (1.4 km apart) and the equal distance from the shore and the proportion of mature individuals within the (100 m). Having established that the two sites were cactus sample were calculated for each trap grid from suYciently homogeneous in microhabitat structure the Wrst 20 (point to nearest plant) trees, shrubs and (see Results), 23 adult N. swarthi (11 females and 12 cacti sampled. males) were radio-tracked between May and July 2002. Abundance data were based on estimates from Cable tie radio-collars (Biotrack Ltd, Dorset, UK), 10 days of trapping on the eight grids. Each grid was weighing less than 5% of each ’s body weight composed of one collapsible Tomahawk trap (single (Kenward 2001) were Wtted under anaesthesia (halo- door rat trap Model 201: 406.4£127£127 mm, Toma- thane). were tracked on foot using TR4 hawk Live Trap, Tomahawk, WI, USA) per point at receivers (Telonics Inc. AZ, USA) and hand-held, Xex- 30 m spacing in a 7£7 array. Abundance was estimated ible three-element Yagi antennas (Biotrack Ltd, Dor- using the Minimum Number Known Alive (MNKA) set, UK). Of this sample, 11 N. swarthi were tracked on method of Krebs (Krebs 1966). MNKA was chosen as the low density R. rattus grid and 12 N. swarthi were N (abundance) was less than 25 for some grids, failing tracked on the high density R. rattus grid. Radio-track- criteria required by closed population estimators based ing shifts alternated between 1600–0200 and 2200–0800 on capture probabilities (Otis et al. 1978). The live- with a Wx obtained for each of four animals every hour. 123 Oecologia

Only four animals were tracked in any one night and the presence of negative spatial association. Each trap data were collected over a total of 26 nights. All analy- (of 49 per grid) was assigned to one of four categories ses were carried out using programme Ranges6 (Ken- based on its history of occupancy: “N. swarthi only”, ward et al. 2003). Separate incremental analyses were “R. rattus only”, “both species” or “neither species” carried out to check that the sample size (number of over the 10 days. Fishers exact tests were used to test Wxes) was suYcient for each animal. When delineating for interspeciWc spatial segregation as some expected the home range, high use cores may be separable from values in the contingency tables were below 5 (Field peripheral areas that are seldom visited (Burt 1943; 2000). Kenward et al. 2001). To do this, we used cluster analy- sis with the single inclusive convex polygon method. Microhabitat Cluster analysis is a linkage method that clusters track- ing Wxes using nearest neighbour distances to form high Spool-and-line tracking was used to assess the micro- use cores (Kenward et al. 2001). Inspection of area habitat selection of N. swarthi and R. rattus. This plots at increasing core percentages (5% increments) method involves attaching a bobbin of thread (spool) for each subject enabled selection of the high use core to an animal the end of which is attached to vegetation by the inXection in the curve of area versus percentage or a rock at the release point. The spools unravel from of locations. The 100% minimum convex polygon the inside, so that as animals move through vegetation (MCP) estimate of home range area (i.e. the estimate the thread plays out without resistance, attaching to before truncation) was also included. This estimate vegetation and rocks or lava (Boonstra and Craine includes movements made by the animals out of the 1985). The trail can then be followed the next day and “home range” or high use core. The 100% MCP compared to the data from a control line oriented at a method was used to obtain the range span (R/span). random angle from the start point to quantify micro- All home range parameters were compared between habitat selection (Cox et al. 2000). This technique is high and low black rat areas. To assess temporal activ- simple, cheap and permits very Wne scale analysis of ity patterns, movement between subsequent hourly behaviour that would not be possible using radio- Wxes indicted activity in that hour. To plot and analyse telemetry or trapping (e.g. Boonstra and Craine 1985; activity patterns on the high and low black rat areas the Key and Woods 1996; Dennis 2002, 2003). In further proportion of animals with at least one active Wx at a support, a recent study has shown that the body mass, particular time interval was calculated for each time survival and trappability of kangaroo rats (Dipodomys interval between 1700 and 0800. Data were pooled spectabilis) is not biased by spool-and-line tracking and across nights by classifying each individual as active for anecdotal observations suggest that behaviour is a particular time interval if there was movement unaVected (Steinwald et al. 2006; M. Steinwald, per- between the two hourly Wxes on any of the nights it was sonal communication). We conducted pilot trials to tracked. Not all individuals were tracked during all reWne our technique and found that attaching the intervals. The proportion of active individuals was then spools to the rump of R. rattus and between the shoul- calculated for each time interval. As the data for each der blades of N. swarthi prevented them chewing them time interval were therefore based on multiple animals oV. Allowing suYcient time for the adhesive to dry also tracked on multiple nights, the data were a mixture of minimised the probability of spool loss. Prompt return repeated measures and independent data. Conse- to the spooling site for spool line analysis minimised quently, a separate 2£2 contingency table was analy- the likelihood of line breakage (probably caused by sed for each hour to compare activity between the high goats in most cases) which was distinguished from line and low R. rattus areas. We also tested for correlation termination by fraying of the cotton at the break point. of nocturnal activity between the high and low R. rattus In most cases the line reattached close to the break area and temporal variation in activity between these point (see also Key and Woods 1996). sites. Traps were set around dusk (1800) and checked between 2200 and 0200. Spools (quilting cocoons: Nm Spatial distribution 140/2, Nylon size 7, 2.4 g, 155 m. DanWeld Ltd., Leigh, Lancs., UK) were unravelled to an appropriate size Data collected from the live-trapping in 2002 in May and weight to Wt each individual animal. Each spool on the high and low density R. rattus grids described was then wrapped in a casing of electrical tape to pre- above and the later survey between December 2002 vent snagging on vegetation. Final spool mass was and January 2003 (the eight grids used to calculate the checked to ensure that it did not exceed 5% of the abundance–habitat correlations) were used to test for body weight of the animal. This guideline is usually 123 Oecologia adhered to during radio-tracking studies to minimise results conWrm that N. swarthi is restricted to one pop- the risk of aVecting the animal’s activity and behavio- ulation in Opuntia and Bursera thorn scrub habitat ural patterns (Kenward 2001). Spools were attached to within the Arid Zone on the north-central coast of San- animals using cyanoacrylate glue (gel form). This adhe- tiago (Fig. 2). In contrast, R. rattus were present at all sive has been tried and tested (Key and Woods 1996) survey sites. and our pilot trials demonstrated that the spool case sloughs oV shortly after the tracking without damage to Habitat associations the underlying skin. The end of the spool was tied to vegetation and the start position marked with Xagging The presence of N. swarthi at the hair tube survey sites tape. Spool lines were analysed the following morning. was signiWcantly correlated with the presence of mature Only lines of minimum length 30 m were analysed. The cacti (rs(15)=0.577, P=0.024). All N. swarthi-occupied Wrst 10 m was classiWed as a “Xight response” and was sites (nine of nine) contained mature cacti and cactus not included in the analysis (Cox et al. 2000). Micro- density was estimable at Wve of these sites [mean cactus habitat variables were recorded over 5 m sections of density: 0.00239/m2§0.000622 (1 SE)]. Individual cacti line. The proportion of sections containing each of the at the other sites were too wide-spaced for accurate following dominant shrub species; C. molle, C. galapa- measurement in the rugged terrain (see Materials and geia, L. peduncularis and Scutia spicata, each of the two methods). Just two of the six unoccupied sites con- dominant tree species; B. graveolens and C. scouleri, tained mature cacti and only one of these sites (north of and mature (>25 pads) and immature (0–25 pads) O. James Bay) contained enough cacti for density estima- galapageia cactus were calculated. Trees and cacti were tion (0.000491/m2). The density of shrubs and trees did included in a given segment if the spool line passed not diVer between occupied and unoccupied sites. under the canopy and/or within 2 m of the trunk. However three of the unoccupied sites, those in the Finally, the proportional occurrence of broken lava Sullivan Bay area, were mainly composed of barren was indexed by observation of 2 m radius plots at 10 m lava with very little vegetation of any kind. As a result, intervals within which the lava was coded as Xat or bro- shrub density was not measurable at any of these sites ken/cracked. This analysis process was repeated along and tree density was estimable at just one of these sites a straight line of equal length running in a random (closest to La Bomba). The species composition of direction from the point of spool line attachment. This shrubs and trees on occupied and unoccupied sites was random line represented the microhabitat availability very diVerent precluding statistical analysis. However, for that animal. Data were collected for 41 N. swarthi and 11 R. rattus. Data were screened for normality and microhabitat components extracted by factor analysis. A GLM (SPSS GLM > Univariate; SPSS v. 11) was then used to Wt models to test if the diVerence between selected and random microhabitat on factor scores was statistically signiWcant within and between species. The individual identity of rats was entered into these mod- els as a blocking factor. These data were collected from six of the trapping grids throughout 2003.

Results

Current distribution

The hair tube survey identiWed ten sites that were inhabited by N. swarthi (Fig. 1). There was some initial uncertainty regarding hair classiWcation at two sites (broken hairs) and hair tube tapes became damp and non-adhesive at a further two sites. However, post-hoc trapping at these sites as part of a study by CDRS per- sonnel conWrmed our predictions from the hair tube sampling (CDRS, unpublished data). The survey Fig. 2 Opuntia and Bursera thorn scrub habitat at La Bomba 123 Oecologia the dominant shrub at occupied sites was C. molle (six plants (P=0.224) and mixed shrub plants (P=0.460) did of nine sites) and at unoccupied sites C. galapageia not diVer between grids. Mean tree size (DBH) did not (two of four sites) with complete absence of C. molle. diVer between grids (B. graveolens: P=0.147 and C. The tree species B. graveolens and C. scouleri were the scouleri: P=0.550) and tree species composition was dominant trees at seven of nine occupied sites and two similar (P=0.686). Cactus age composition was also of three unoccupied sites. At La Bomba the abundance comparable (P=0.204). of N. swarthi was positively correlated with cactus den- The home range area, MCP area and range span of sity (r(6)=0.962, P=0.0001) and the proportion of males and females respectively (Table 1) were not sig- mature cacti (r(6)=0.765, P=0.027) but not with shrub niWcantly diVerent on high versus low R. rattus grids density (r(6)=0.689, P=0.059). Of dominant plant spe- (P>0.1 for all analyses using Mann Whitney U tests). cies, N. swarthi abundance was positively correlated The proportion of active N. swarthi was signiWcantly with L. peduncularis (r(6)=0.880, P=0.004) and nega- higher on the high R. rattus density grid at 0600 (sun- tively correlated with C. galapageia (r(6) = ¡0.674, rise) and marginally higher in the early evening, one P=0.067) and B. graveolens (r(6) = ¡0.781 P=0.022). A hour prior to dusk, at 1700 (2=7.213, df=1, P=0.007 correlation matrix of the signiWcant habitat variables and 2=2.738, df=1, P=0.098; Fig. 4). The variance showed that cactus density was strongly positively cor- structure of activity diVered between grids (Levene’s related with the proportion of mature cacti, shrub den- test: F=5.640, df=1, 30, P=0.024) and activity patterns sity and with proportion of L. peduncularis and were not correlated (r(14)=0.493, P=0.053). This strongly negatively correlated with C. galapageia and appears to be due to the greater Xuctuation in the pro- B. graveolens (all r>0.7). The abundance of R. rattus portion of active individuals during the latter half of was not correlated with any of the microhabitat vari- the night/early morning on low R. rattus density grids ables or with the abundance of N. swarthi. However, it compared with the higher, sustained proportion of was clear that the ratio of N. swarthi to R. rattus on the coastal grids (grids 1–4, Fig. 3), greatly exceeded the ratio inland which was closer to unity (grids 5–8, Table 1 Home range area by mononuclear clusters truncated to % core (HR area), home range area by 100% minimum convex Fig. 3). polygon method (MCP) and range span (R/span) of the 100% MCP for animals tracked on a low and high R. rattus density grid Spatio-temporal activity R. rattus N. swarthi N % Core HR MCP R/span density sex area (ha) (ha) (m) The structure and composition of the habitat did not diVer between the high (42 R. rattus:66 N. swarthi) and Low Male 42 100 4.67 4.67 319 low (8 R. rattus:79 N. swarthi) black rat density areas 18 100 1.5 1.5 173 W 26 95 0.12 0.27 121 con rming that the comparison of N. swarthi activity 33 95 0.63 1.12 129 would not be biased by variation in measurable habitat 44 100 4.26 4.26 339 variables. There was no signiWcant diVerence in the 42 90 1.57 4.07 376 index of C. molle (P=0.234), mixed shrub (P=0.509) 40 100 3.05 3.05 242 Median 1.57 3.05 242 tree (P=0.380) and O. galapageia (P=0.677) density. In High Male 31 100 1.97 1.97 223 addition, the mean size (volume) of individual C. molle 48 100 1.55 1.55 218 58 95 3.12 9.56 528 39 100 0.35 0.35 73 32 95 0.58 1.13 144 Median 1.55 1.55 218 Low Female 27 95 0.22 0.97 236 31 100 0.24 0.24 63 43 95 0.65 2.13 229 20 95 1.38 2.44 296 Median 0.45 1.55 233 High Female 41 95 0.19 0.31 90 35 100 0.67 0.67 130 35 95 0.34 0.68 111 38 95 0.85 1.29 189 28 100 0.53 0.53 100 23 95 0.48 1.13 232 16 100 0.41 0.41 79 Fig. 3 Abundance (MNKA) of N. swarthi and R. rattus from live- Median 0.48 0.67 111 trapping on four coastal (1–4) and four inland (5–8) grids at the start of the 2003 wet season (December 2002–January 2003) N No. of Wxes 123 Oecologia

P=0.004) but use of the remaining microhabitat types was consistent with random. However, while it is likely that both rodent species use similar habitats, R. rattus is a subset (Fig. 5). This is probably due to the smaller sample of spooled R. rattus. The relative selection for each microhabitat did not diVer between species (com- ponent 1: F=2.395, df=1, 50, P=0.128; component 2: F=0.060, df=1, 50, P=0.807; component 3: F=0.040, df=1, 50, P=0.843) and interspeciWc similarity in micro- habitat use is evident upon inspection of the ordination plot (Fig. 5). Fig. 4 The proportion of active individuals (N. swarthi) at hourly intervals (1700–0800) Spatial distribution active animals on high R. rattus grids throughout the The distribution of species between trap types (R. rat- night. tus only, N. swarthi only, R. rattus and N. swarthi or neither species) in 2002 (two grids) and 2003 (seven Microhabitat selection grids) was not signiWcantly diVerent from random indi- cating that there was no signiWcant negative (or posi- Three components were extracted by factor analysis tive) spatial association between individuals of the two (eigenvalues > 1) which together explained 52.43% of species (Table 3). the variation in the microhabitat dataset as revealed by spool-and-line analysis (Table 2). Analysis of the paired (chosen versus random) data Discussion revealed that N. swarthi selected microhabitat type/ component 2 (mature cacti and shrubs with broken Our survey results reveal that the world population of lava) (F=31.274, df=1, 40, P<0.0001) and avoided N. swarthi is now restricted to a 14 km strip on the microhabitat type/component 1 (trees and juvenile north-central coast of Santiago. There was no evidence cacti) (F=8.652, df=1, 40, P=0.005). Use did not diVer from availability for component 3 (mixed shrubs). Results are therefore consistent with the important plant species identiWed in the previous analyses. Simi- larly, R. rattus exhibited signiWcant selection for microhabitat type/component 2 (F=14.078, df=1, 10,

Table 2 Factor loadings for microhabitat variables within each component (eigenvalues > 1) Plant species/lava type Component

123

Bursera graveolens 0.671 ¡0.178 ¡0.075 Croton scouleri 0.721 0.159 0.303 Opuntia galapageia (juvenile) 0.595 ¡0.001 ¡0.531 Opuntia galapageia (adult) 0.381 0.543 ¡0.355 Clerodendrum molle ¡0.046 0.483 0.545 Castela galapageia 0.470 ¡0.429 0.518 Lantana peduncularis 0.264 0.485 0.208 Scutia spicata 0.083 ¡0.409 0.370 Lava complexity ¡0.103 0.690 0.204 Microhabitat types can be described as component 1, Trees (B. graveolens and C. scouleri) with C. galapageia and immature cacti; component 2, Shrubs (C. molle and L. peduncularis) with Fig. 5 Microhabitat use. Ordination plot of chosen spool route mature cacti and broken lava and component 3, Shrubs (C. molle, microhabitat data for the Wrst two microhabitat components for C. galapageia and S. spicata) N. swarthi (open symbols) and R. rattus (closed symbols) 123 Oecologia

Table 3 Contingency tables for trap occupancy in 2002 (two density is lower (e.g. Trombulak 1985; Dickman grids) and 2003 (eight grids) 1986). Conversely, if resource competition predomi- 2002, coast (grid 2) N. swarthi + N. swarthi ¡ P nates then we may expect home range to be smaller R. rattus +6 1 where R. rattus density is lower to reXect increased R. rattus ¡ 38 4 0.554 resource availability (e.g. Taitt 1981; Sullivan et al. 2002, coast (grid 4) N. swarthi + N. swarthi ¡ 1983). However, home range delineation revealed no R. rattus +278 V R. rattus ¡ 9 5 0.476 di erence in the range size of N. swarthi in the high 2003, coast (grid 1) N. swarthi + N. swarthi ¡ density R. rattus area compared to the low density R. R. rattus +9 1 rattus area. R. rattus ¡ 38 1 0.370 Nevertheless, radio-tracking of N. swarthi on the 2003, coast (grid 2) N. swarthi + N. swarthi ¡ R. rattus +5 1 high R. rattus density area revealed an intriguing trend R. rattus ¡ 42 1 0.232 towards increased activity levels and duration (includ- 2003, coast (grid 3) N. swarthi + N. swarthi ¡ ing pre-dusk and post-dawn periods) compared to the R. rattus +271 low density R. rattus area. For comparison, a sample of R. rattus ¡ 21 0 1.000 2003, coast (grid 4) N. swarthi + N. swarthi ¡ 54 R. rattus radio-tracked in Hawaii showed peak activ- R. rattus +340 ity between 2100 and 0300 (Tobin et al. 1996). As the R. rattus ¡ 15 0 Test invalida activity pattern of R. rattus is believed to be fairly gen- 2003, inland (grid 5) N. swarthi + N. swarthi ¡ eral it is likely that the species exhibits a similar activity R. rattus +136 R. rattus ¡ 20 10 1.000 pattern in the Galápagos (Barnett et al. 1975). N. swar- 2003, inland (grid 6) N. swarthi + N. swarthi ¡ thi on the low R. rattus grid show a comparable activity R. rattus +1414 peak between approximately 2000 and 0300. However, R. rattus ¡ 12 9 0.774 on the high R. rattus grid they may be exploiting the 2003, inland (grid 7) N. swarthi + N. swarthi ¡ R. rattus +9 20 low activity phase of R. rattus around dusk and dawn. R. rattus ¡ 7131.000This may reXect an increase in foraging eVort necessary 2003, inland (grid 8) N. swarthi + N. swarthi ¡ to compensate for the mutual exploitation of limited R. rattus +177 resources or may be a response to disturbance or inter- R. rattus ¡ 20 5 0.520 ference by black rats in the vicinity. In natural desert a Grid 4, 2003 analysis was invalid as all traps were occupied by rodent communities, the additional energy cost of N. swarthi extended foraging time has been shown to outweigh the beneWts resulting in reduced foraging activity of N. swarthi at, or in the vicinity of, Sullivan Bay, the (Mitchell et al. 1990). The cost–beneWt balance is area from which the type specimens were collected in uncertain here but there may be negative implications 1906 (Orr 1938). In contrast, R. rattus has certainly col- for the Wtness of N. swarthi on high density R. rattus onised every Galápagos habitat from coastal desert to grids. montane forest (Clark 1980) and our Wndings indicate Finally, presuming that overlap in space and time that Santiago is no exception. These results therefore correlates with the probability of encounter between constitute the Wrst evidence of N. swarthi range con- the dominant R. rattus and subordinate N. swarthi we traction since R. rattus introduction, substantiating the might expect there to be selection for microhabitat seg- need to analyse the N. swarthi–R. rattus relationship at regation. However, there were no interspeciWc diVer- the last apparent stronghold. ences in microhabitat use. In fact, both species strongly Given that R. rattus is larger, more aggressive and selected the same microhabitat type; areas containing behaviourally dominant to N. swarthi (D.B. Harris, large mature cacti, the typically expansive, dense shrub unpublished data) it is surprising that interspeciWc den- species C. molle and L. peduncularis and broken lava. sities were unrelated. Nevertheless, Hutchinson (1961) This microhabitat type is found on the coast and is pre- proposed that even under strong competition spatial sumably ideal for rodents as it is composed of thick and/or temporal heterogeneity could promote species shrub cover and complex lava topography that coexistence. However, the investigation of interspeciWc together oVer good protection from predators (hawks spatial overlap demonstrated that the two species and owls) and relatively high food availability (shrubs intermix freely in space. Under such conditions we and cacti). may expect the home range size of the subordinate N. Taken together these results suggest that the appar- swarthi to vary with the density of R. rattus. For exam- ent coexistence of N. swarthi and R. rattus is not facili- ple, if interspeciWc competition is important then we tated by spatial or temporal partitioning and that the may expect the home range to be larger where R. rattus two species have a high probability of encounter during 123 Oecologia normal activity. In other words, the data suggest that In conclusion, there is no obvious spatial segrega- the species do not compete for space in the last strong- tion and N. swarthi activity did not vary with R. rattus hold where N. swarthi and R. rattus have been sympat- density. However, the falsiWcation of our initial ric up to 400 years (Patton et al. 1975; Morris 1983). By hypotheses through this preliminary exploration of comparison, R. rattus and N. indefessus were sympatric pattern does not eliminate the premise of competitive for just 4 years on Santa Cruz Island before the decline coexistence in this system. Temporal variation in and extinction of the latter (Clark 1984). This anec- resource availability with intra- and inter-annual cli- dote, considered together with the apparent, and rela- matic Xuctuation may provide the axis of environmen- tively recent (since 1906) loss of N. swarthi from tal heterogeneity necessary for species coexistence Sullivan Bay, prompts us to ask: what is special about (Kotler and Brown 1988). It has been demonstrated the Wnal stronghold on the north-central coast of Santi- that food-limited R. rattus populations can undergo ago? extreme Xuctuations in density in Galápagos thorn Our habitat investigations revealed a strong corre- scrub (Clark 1980). This almost certainly leads to lation between the occurrence of mature Opuntia cac- occasional local extinction during dry periods in the tus and the presence of N. swarthi. Indeed, the region Galápagos arid zone (Clark 1980). Importantly, this occupied by rice rats appears to be unique in its habi- seems particularly feasible on the exceptionally arid, tat composition with a high density of mature Opuntia rain-shadowed north coast of Santiago (D.B. Harris, cactus. Furthermore, within this region, the abun- unpublished data). This periodic respite from interfer- dance of N. swarthi was correlated with cactus density ence by the larger R. rattus may be suYcient to allow and the proportion of mature cacti. This may reXect a coexistence of the two species without the need for N. positive relationship between consumer (N. swarthi) swarthi to adjust its space use, habitat preferences and and preferred resources in this chosen habitat (Rosen- activity patterns. The other vital part of any mecha- zweig 1991). The importance of Opuntia cactus to N. nism of coexistence is a trade-oV between the abilities swarthi is further supported by a lack of N. swarthi of the competitors to utilise diVerent parts of the axis reproduction on inland Arid Zone (low cactus) com- (Kotler and Brown 1988). In this case the trade-oV pared with coastal Arid Zone (high cactus) grids dur- may be based on interference competition, resulting in ing a dry year (D.B. Harris, unpublished data). In behavioural dominance by the larger, aggressive R. contrast to the emerging N. swarthi–Opuntia relation- rattus, perhaps with priority access to the most pre- ship, R. rattus population density did not correlate ferred resources during the wet, resource rich season with any of the main habitat components identiWed at while the native N. swarthi may be the superior, or La Bomba. The black rat is renowned for ecological exclusive, exploiter of the locally abundant endemic Xexibility which at least partly explains its success as a cactus, as suggested by the high N. swarthi to R. rattus widespread invader (Clark 1980, 1981; Lehtonen et al. ratio on the coastal trap grids (Fig. 3). The fruits of the 2001; Courchamp et al. 2003; Russell and Clout 2004). Opuntia cactus are plentiful in the wet season and the This ecological plasticity coupled with behavioural succulent cladodes are available year round. Superior dominance over N. swarthi may have given the black or exclusive exploitation of cactus may explain why N. rat the competitive edge in the Sullivan Bay area swarthi is able to maintain stable population levels which is composed almost entirely of lava with very lit- throughout the dry season when availability of alter- tle vegetation. It is feasible that competition is intensi- native resources is low (D.B. Harris, unpublished Wed in such habitat where interspeciWc encounter rate data). Alternatively this trade-oV may be described by and/or resource overlap are likely to be higher. This resource partitioning as a consequence of exploitation may have led to the local displacement of N. swarthi. competition. Further research is needed to deWne the However, N. swarthi still appear able to occupy some resource axis of environmental heterogeneity and to suboptimal habitat in the presence of R. rattus. Within distinguish between the alternative mechanisms of the Arid Zone at La Bomba, N. swarthi occurs in areas interference and exploitation. inland from the high density cactus core or probable “source habitat”. However, its densities in these sub- Hypotheses for future testing optimal habitats are notably lower (Fig. 3) and as mentioned, reproduction may cease under certain We hypothesise that the coexistence of N. swarthi and conditions suggesting that such low quality peripheral R. rattus is facilitated by temporal variation in areas might act as “sink habitat” (Pulliam 1988; resource availability (e.g. Ben-Natan et al. 2004) cou- Pulliam and Danielson 1991; D.B. Harris, unpublished pled with a diVerence in resource use which may have data). evolved to reduce competition and/or may be dictated 123 Oecologia by morphological or physiological feeding constraints areas where our hair tube results were uncertain and Paul John- (e.g. Jenkins and Ascanio 1993; Begon et al. 1996). As ston for statistical advice. Finally, we would like to thank Felipe Cruz and the Project Isabela team for their logistic support. The the presence or absence, microhabitat selection and Weld work described within complies with the current laws of the abundance patterns of N. swarthi are all closely corre- country in which it was performed (Ecuador). lated with Opuntia cactus density, we speculate that cactus might be a crucial resource refuge for N. swarthi. Indeed, if N. swarthi were the superior exploiter or had References exclusive access to Opuntia resources, then coexistence may be possible despite the likely costs of behavioural Abramsky Z, Dyer MI, Harrison PD (1979) Competition among sub-ordinance to aggressive interference (Keddy 2001; small mammals in experimentally perturbed areas of the shortgrass prairie. Ecology 60:530–536 D.B. Harris, unpublished data). This hypothesis should Abramsky Z, Rosenzweig ML, Pinshow B, Brown JS, Kotler BP, direct future research towards a study of interspeciWc Mitchell WA (1990) Habitat selection: an experimental test diet relations across seasons, with particular emphasis with two gerbil species. Ecology 71:2358–2369 on the role of the cactus as a resource refuge. Interest- Alvarez VB, Gonzalez AC (1991) The critical condition of hutias in Cuba. Oryx 25:206–208 ingly, preliminary captive observations show that indi- Barnett SA, Cowan PE, Prakash I (1975). Circadian rhythm of vidual R. rattus either will not, or are unable to, movements of the house rat, Rattus rattus L. Indian J Exp consume cactus fruits even when the fruits are opened Biol 13:153–155 and pulp exposed. Furthermore, while N. swarthi regu- Begon M, Harper JL, Townsend CR (1996) Ecology: individuals, populations and communities, 3rd edn. Blackwell, Oxford larly climb and forage in the cactus canopy, we have no Bender EA, Case TJ, Gilpin ME (1984) Perturbation experi- evidence for ascent of cacti by R. rattus (D.B. Harris ments in community ecology: theory and practice. Ecology and S.D. Gregory, personal observation). 65:1–13 It is important to note that implicit within our coex- Ben-Natan G, Abramsky Z, Kotler BP, Brown JS (2004) Seeds redistribution in sand dunes: a basis for coexistence of two istence hypothesis is the assumption that the two spe- rodent species. Oikos 105:325–335. DOI 10.1111/j.0030- cies compete. The historical pattern of Galápagos 1299.2004.12948.x rodent certainly suggests that R. rattus is Boonstra R, Craine ITM (1985) Natal nest location and small reducing the survival of N. swarthi in its last apparent mammal tracking with a spool and line technique. Can J Zool 64:1034–1036 stronghold. However, there is no unequivocal, scien- Bowers MA (1982) Foraging behavior of Heteromyid rodents: tiWc evidence to support our postulation. We therefore Weld evidence of resource partitioning. J Mammal 63:361– recommend the use of a replicated “press experiment” 367 (sensu Bender et al. 1984). This would involve repres- Bowers MA, Thompson DB, Brown JH (1987) Spatial organisa- R. rattus tion of a desert rodent community: food addition and species sion of density in experimental plots. The removal. Oecologia 72:77–82 demographic response of individual N. swarthi to R. Brosset A (1963) Statut actuel des mammifères des îles Galapa- rattus density repression should then be monitored gos. Mammalia 27:323–340 with a focus on vital rates such as survival and fecun- Bryce J, Johnson PJ, Macdonald DW (2002) Can niche use in red and grey squirrels oVer clues for their apparent coexistence? dity (Krebs 1995; Begon et al. 1996; Eccard and Ylonen J Appl Ecol 39:875–887. DOI 10.1046/j.1365- 2003). Any impact of R. rattus on N. swarthi should 2664.2002.00765.x then be compared in low versus high density cactus Burt WH (1943) Territoriality and home range concepts as ap- areas to test for habitat-dependent competition which plied to mammals. J Mammal 24:346–352 Clark DA (1981) Foraging patterns of black rats across a desert- may provide further support for the proposed mecha- montane forest gradient in the Galápagos Islands. Biotropica nism of competition. We predict that competition will 13:182–194 be more severe in low density cactus habitat. Clark DA (1984) Native land mammals. In: Perry R (ed) Key In the meantime, a thorough search of high density environments: Galápagos. Pergamon Press, Oxford, pp 225– 231 cactus patches on other islands may reveal further Clark DB (1980) Population ecology of Rattus rattus across a des- extant populations of “extinct” Galápagos rodents. ert-montane forest gradient in the Galápagos Islands. Ecol- ogy 61:1422–1433 Courchamp F, Chapuis JL, Pascal M (2003) Mammal invaders on Acknowledgements This research was funded by grants from islands: impact, control and control impact. Biol Rev 78:347– the Galápagos Conservation Trust, Flora and Fauna Interna- 383. DOI 10.1017/S1464793102006061 tional, Columbus Zoo and the Peoples Trust for Endangered Spe- Cox MPG, Dickman CR, Cox WG (2000) Use of habitat by the cies together with support from the James Teacher Memorial black rat (Rattus rattus) at North Head, New South Wales: an Trust. We thank the staV of the Charles Darwin Research Station observational and experimental study. Austral Ecol 25:375– and the Galápagos National Park Service, especially Brand Phil- 385. DOI 10.1046/j.1442-9993.2000.01050.x lips and Brian Cooke, for their collaboration and logistic support. Dennis AJ (2002) The diet of the musky rat-kangaroo, Hyp- We are also grateful to Amie IllWeld for Weld assistance, Gillian siprymnodon moschatus, a rainforest specialist. Wildl Res Key and Marjorie Riofrio for adapting their study to incorporate 29:209–219. DOI 10.1071/WR00052 123 Oecologia

Dennis AJ (2003) Scatter-hoarding by musky rat-kangaroos, Ligtvoet W, Van Wijngaarden A (1994) The colonization of the Hypsiprymnodon moschatus, a tropical rain-forest marsupial island of Noord-Beveland (The Netherlands) by the com- from Australia: implications for seed dispersal. J Trop Ecol mon vole Microtus arvalis, and its consequences for the root 19:619–627. DOI 10.1017/S0266467403006023 vole M. oeconomus. Lutra 37:1–28 Dickman CR (1986) An experimental manipulation of the inten- Lofgren O (1995) Spatial organization of cyclic Clethrionomys fe- sity of interspeciWc competition: eVects on a small marsupial. males: occupancy of all available space at peak densities? Oi- Oecologia 70:536–543 kos 72:29–35 Dowler RC, Carroll DS, Edwards CW (2000) Rediscovery of ro- Macdonald DW, Bryce JM, Thom MD (2001). Introduced mam- dents (Genus Nesoryzomys) considered extinct in the Galá- mals: do carnivores and herbivores usurp native species by pagos Islands. Oryx 34:109–117. DOI 10.1046/j.1365- diVerent mechanisms? In: Pelz HJ, Cowan DP, Feare CJ 3008.2000.00104.x (eds) Advances in vertebrate pest management II. Filander Eccard JA, Ylonen H (2003) InterspeciWc competition in small Verlag, Fürth, pp 11–44 rodents: from populations to individuals. Evol Ecol 17:423– McMullen CK (1999) Flowering plants of the Galápagos. Cornell 440. DOI 10.1023/A:1027305410005 University Press, Ithaca Field A (2000) Discovering statistics using SPSS for Windows. Meehan AP (1984) Rats and mice their biology and control. SAGE Publications Ltd, London Rentokil Limited, East Grinstead, W.Sussex Greenwood JJD (1996) Basic techniques. In: Sutherland WJ (ed) Mitchell WA, Abramsky Z, Kotler BP, Pinshow B, Brown JS Ecological census techniques: a handbook. Cambridge Uni- (1990) The eVect of competition on foraging activity in des- versity Press, Cambridge, pp 11–110 ert rodents: theory and experiments. Ecology 71:844–854 Gurnell J, Wauters LA, Lurz PWW, Tosi G (2004) Alien species Monamy V, Fox BJ (1999) Habitat selection by female Rattus lu- and interspeciWc competition: eVects of introduced eastern treolus drives asymmetric competition and coexistence with grey squirrels on red squirrel population dynamics. J Anim Pseudomys higginsi. J Mammal 80:232–242 Ecol 73:26–35. DOI 10.1111/j.1365-2656.2004.00791.x Morris D (1983) Field tests of competitive interference for space Hutchinson GE (1961) The paradox of the plankton. Am Nat among temperate-zone rodents. Can J Zool 61:1517–1523 95:137–145 Morris DW (1987) Ecological scale and habitat use. Ecology Innes JG (2005) Ship rat. In: King CM (ed) The handbook of New 68:362–369 Zealand mammals, 2nd edn. Oxford University Press, South Niethammer J (1964) Contribution a la connaissance des mammi- Melbourne feres terrestres de l’ile indefatigable (Santa Cruz), Galapa- Jackson MH (1993) Galápagos: a natural history. Second revised gos. Resultats de l’Expedition Allemande aux Galapagos, and expanded edition. University of Calgary Press, Calgary 1962/63. Mammalia 28:593–606 Jenkins SH, Ascanio R (1993) A potential nutritional basis for re- Orr RT (1938) A new rodent of the genus Nesoryzomys from the source partitioning by desert rodents. Am Midl Nat 130:164–172 Galapagos Islands. In: Proceedings of the California Acad- Jones M, Mandelik Y, Dayan T (2001) Coexistence of temporally emy of Science, 4th series, vol 23, pp 303–306 partitioned spiny mice: roles of habitat structure and forag- Otis DL, Burnham KP, White GC, Anderson DR (1978) Statisti- ing behavior. Ecology 82:2164–2176 cal inference from capture data on closed animal popula- Jorgensen EE, Demarais S (1999) Spatial scale dependence of ro- tions. Wildl Monogr 62:135 dent habitat use. J Mammal 80:421–429 Patton JL, Yang SY, Myers P (1975) Genetic and morphological Keddy PA (2001) Competition, 2nd edn. Kluwer, Dordrecht divergence among introduced rat populations of the Galapa- Kenward RE (2001) A manual for wildlife radio-tagging. Aca- gos Archipelago, Ecuador. Syst Zool 24:296–310 demic, London Price MV (1978) The role of microhabitat in structuring desert ro- Kenward RE, Clarke RT, Hodder KH, Walls SS (2001) Density dent communities. Ecology 59:910–921 and linkage estimators of home range: nearest-neighbor clus- Price MV, Waser NM (1985) Microhabitat use by Heteromyid ro- tering deWnes multinuclear cores. Ecology 82:1905–1920 dents: eVects of artiWcial seed patches. Ecology 66:211–219 Kenward RE, South AB, Walls SS (2003) Ranges6 v1.2: for the Pulliam HR (1988) Sources, sinks, and population regulation. Am analysis of tracking and location data. Online manual. Ana- Nat 132:652–661 track Ltd., Wareham. ISBN 0-9546327-0-2 Pulliam HR, Danielson BJ (1991) Sources, sinks, and habitat Key G, Muñoz Heredia E (1994) Distribution and current status selection: a landscape perspective on population dynamics. of rodents in the Galápagos. Noticias de Galápagos 53:21–25 Am Nat 137:S50–S66 Key GE, Woods RD (1996) Spool-and-line studies on the Rosenzweig ML (1991) Habitat selection and population interac- behavioural ecology of rats (Rattus spp.) in the Galápagos Is- tions: the search for mechanism. Am Nat 137:S5–S28 lands. Can J Zool 74:733–737 Russell JC, Clout MN (2004) Modelling the distribution and Kotler BP, Brown JS (1988) Environmental heterogeneity and interaction of introduced rodents on New Zealand oVshore the coexistence of desert rodents. Annu Rev Ecol Syst islands. Global Ecol Biogeogr 13:497–507. DOI 10.1111/ 19:281–307 j.1466-822X.2004.00124.x Krebs CJ (1966) Demographic changes in Xuctuating populations Schoener TW (1974) Resource partitioning in ecological commu- of Microtus californicus. Ecol Monogr 36:239–271 nities. Science 185:27–39 Krebs CJ (1995) Two paradigms of population regulation. Wildl Shea K, Chesson P (2002) Community ecology theory as a frame- Res 22:1–10 work for biological invasions. Trends Ecol Evol 17:170–176. Krupa JJ, Haskins KE (1996) Invasion of the meadow vole (Mi- DOI 10.1016/S0169-5347(02)02495-3 crotus pennsylvanicus) in southeastern Kentucky and its pos- Snell H, Rea S (1999) The 1997–98 El Niño in Galápagos: can sible impact on the southern bog lemming (Synaptomys 34 years of data estimate 120 years of pattern? Noticias de cooperi). Am Midl Nat 135:14–22 Galápagos 60:11–20 Lehtonen JT, Mustonen O, Ramiarinjanahary H, Niemela J, Rita Stephenson PJ (1993) The small mammal fauna of Reserve Spec- H (2001) Habitat use by endemic and introduced rodents iale d’Analamazaotra, Madagascar: the eVects of human dis- along a gradient of forest disturbance in Madagascar. Biodiv- turbance on endemic species diversity. Biodivers Conserv ers Conserv 10:1185–1202. DOI 10.1023/A:1016687608020 2:603–615

123 Oecologia

Steinwald MC, Swanson BJ, Waser PM (2006) The eVects of Trombulak SC (1985) The inXuence of interspeciWc competition spool-and-line tracking on small desert mammals. Southwest on home range size in chipmunks (Eutamias). J Mammal Nat (in press) 66:329–337 Sullivan TP, Sullivan DS, Krebs CJ (1983) Demographic respons- Yom-Tov Y, Yom-Tov S, Moller H (1999) Competition, coexis- es of a chipmunk (Eutamias Townsendii) population with tence, and adaptation amongst rodent invaders to PaciWc and supplemental food. J Anim Ecol 52:743–755 New Zealand Islands. J Biogeogr 26:947–958. DOI 10.1046/ Taitt MJ (1981) The eVect of extra food on small rodent popula- j.1365–2699.1999.00338.x tions. 1. Deermice (Peromyscus maniculatus). J Anim Ecol Zavaleta E, Hobbs R, Mooney H (2001) Viewing invasive species 50:111–124 removal in a whole-ecosystem context. Trends Ecol Evol Teerink BJ (1991) Hair of west-European mammals. Cambridge 16:454–459. DOI 10.1016/S0169-5347(01)02194-2 University Press, Cambridge Ziv Y, Abramsky Z, Kotler BP, Subach A (1993) Interference Tobin ME, Sugihara RT, Koehler AE, Ueunten GR (1996) Sea- competition and temporal and habitat partitioning in two sonal activity and movements of Rattus rattus (Rodentia, gerbil species. Oikos 66:237–246 Muridae) in a Hawaiian macadamia orchard. Mammalia 60:3–13

123